Diabetes Care Volume 40, August 2017 1065

Frans K. Gorus,1,2 Eric V. Balti,1,2 Twenty-Year Progression Rate to Anissa Messaaoui,3 Simke Demeester,1,2 Annelien Van Dalem,1,2 Olivier Costa,1,2 Clinical Onset According to Harry Dorchy,3 Chantal Mathieu,4 fi Luc Van Gaal,5 Bart Keymeulen,1,6 Autoantibody Pro le, Age, and Daniel¨ G. Pipeleers,1 and Ilse Weets,1,2 for HLA-DQ Genotype in a Registry- the Belgian Diabetes Registry* Based Group of Children and Adults With a First-Degree Relative With Type 1 Diabetes Diabetes Care 2017;40:1065–1072 | https://doi.org/10.2337/dc16-2228

OBJECTIVE We investigated whether islet autoantibody profile, HLA-DQ genotype, and age influenced a 20-year progression to diabetes from first autoantibody positivity (autoAb+)infirst-degree relatives of patients with type 1 diabetes. PATHOPHYSIOLOGY/COMPLICATIONS 1Diabetes Research Center, Vrije Universiteit RESEARCH DESIGN AND METHODS Brussel, , Belgium Persistently islet autoAb+ siblings and offspring (n =462)under40yearsofagewere 2Department of Clinical Chemistry, Universitair followed by the Belgian Diabetes Registry. AutoAbs against insulin (IAA), GAD Ziekenhuis Brussel, Brussels, Belgium 3Department of Diabetology, Hopitalˆ Universitaire (GADA), IA-2 antigen (IA-2A), and zinc transporter 8 (ZnT8A) were determined by des Enfants Reine Fabiola, Brussels, Belgium radiobinding assay. 4Department of Endocrinology, Universitair Ziekenhuis Leuven, Leuven, Belgium RESULTS 5Department of Endocrinology, Diabetology and The 20-year progression rate of multiple-autoAb+ relatives (n = 194) was higher than Metabolism, Universitair Ziekenhuis Antwerpen, + , Belgium that for single-autoAb participants (n = 268) (88% vs. 54%; P < 0.001). Relatives 6Department of Diabetology, Universitair positive for IAA and GADA (n = 54) progressed more slowly than double-autoAb+ Ziekenhuis Brussel, Brussels, Belgium + individuals carrying IA-2A and/or ZnT8A (n = 38; P = 0.001). In multiple-autoAb Corresponding author: Ilse Weets, ilse.weets@ relatives, Cox regression analysis identified the presence of IA-2A or ZnT8A as the uzbrussel.be. only independent predictors of more rapid progression to diabetes (P < 0.001); in Received 17 October 2016 and accepted 22 April single-autoAb+ relatives, it identified younger age (P < 0.001), HLA-DQ2/DQ8 geno- 2017. type (P < 0.001), and IAA (P = 0.028) as independent predictors of seroconversion This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/ to multiple positivity for autoAbs. In time-dependent Cox regression, younger age suppl/doi:10.2337/dc16-2228/-/DC1. P HLA-DQ2/DQ8 P ( =0.042), genotype ( = 0.009), and the development of additional F.K.G. and E.V.B. contributed equally to this autoAbs (P = 0.012) were associated with more rapid progression to diabetes. work. *A complete list of the Members of the Belgian CONCLUSIONS Diabetes Registry can be found in the Supple- In single-autoAb+ relatives, the time to multiple-autoAb positivity increases with age mentary Data online. and the absence of IAA and HLA-DQ2/DQ8 genotype. The majority of multiple- © 2017 by the American Diabetes Association. autoAb+ individuals progress to diabetes within 20 years; this occurs more rapidly Readers may use this article as long as the work in the presence of IA-2A or ZnT8A, regardless of age, HLA-DQ genotype, and number is properly cited, the use is educational and not for profit, and the work is not altered. More infor- of autoAbs. These data may help to refine the risk stratification of presymptomatic mation is available at http://www.diabetesjournals type 1 diabetes. .org/content/license. 1066 Twenty-Year Progression to Type 1 Diabetes Diabetes Care Volume 40, August 2017

Type 1 diabetes is a chronic autoimmune RESEARCH DESIGN AND METHODS clinical onset. BMI was expressed as an disease with an incidence peak around Participants SD score (BMI-SDS) by comparison with puberty (1,2). Observational studies Between March 1989 and December an age- and sex-matched cohort (15). from births in Germany, Finland, and 2015, the BDR consecutively recruited sib- Colorado have followed children at fa- lings or offspring (under 40 years of age at Analytical Methods The presence of IAA, GADA, IA-2A, and milial or HLA-DQ/DR genotype–inferred study entry) of type 1 probands with di- ZnT8A was determined by liquid-phase disease risk for the development of auto- abetes according to previously defined radiobinding assay (12), and that of HLA- antibodies (autoAbs) against insulin criteria (2). The probands are considered DQ polymorphisms was determined by (IAA), GAD65 (GADA), and IA-2 antigen representative of the Belgian population allele-specific oligonucleotide genotyping (IA-2A) and established that .90% of of patients with type 1 diabetes (2,6). Af- (16), as described previously. cDNAs for multiple-autoAb+ individuals progress to ter obtaining written informed consent the preparation of radioligands by in vitro clinical onset within 20 years from sero- from each relative or their parents, a transcription translation were gifts from conversion, with the first autoAb appear- short questionnaire with demographic, A.˚ Lernmark (when at University of Wash- ing most often before 5 years of age (3). familial, and personal information was ington, Seattle, WA) for full-length 65-kDa Reports that also included adults have, completed at each visit, and blood sam- glutamate decarboxylase, M. Christie however, indicated that an important ples were taken at study entry and, as a (King’s College School of Medicine and fraction of seroconversions to autoAb rule, yearly thereafter. Only relatives with Dentistry, London, UK) for the intracellu- positivity occurs after 10 years of age (4) two or more contacts during the fol- lar portion of IA-2, and J.C. Hutton (Barbara and that the disease can become clini- low-up period (7,029 individuals), with Davis Center for Childhood Diabetes, cally overt at any age, with development the last contact being at diagnosis in the Aurora, CO) for the dimeric hybrid ZnT8A of the disease occurring in adulthood in case of progression to diabetes, were in- construct generated by the fusion of CR the majority of patients (2,5–7). There is cluded in this study. The median intervals and CW (zinc transporter-8 carboxy- growing consensus that immune inter- between successive visits ranged be- terminal constructs carrying, respec- ventions in patients with type 1 diabetes tween 11 and 13 months for the various tively, Arg325 and Trp325). ZnT8A were should concentrate on the presymptom- groups of relatives studied. Diabetes was determined in all relatives positive for atic disease phase (8–10). Relatively non- diagnosedaccordingtotheAmerican IAA, GADA, and/or IA-2A but in only a frac- specific interventions, such as anti-CD3 Diabetes Association criteria (14). tion of participants who tested negative therapy, may have to include adults first, The study was conducted in accordance for the other three autoAbs (n = 546; to evaluate safety and efficacy, before with the guidelines in the Declaration of 8.7%), in view of the reported low preva- enrolling children (10). Consequently, Helsinki, as revised in 2013 (accessed lence of solitary ZnT8A positivity (4,12,17). there is a need to compare criteria for on 18 May 2017; https://www.wma disease staging (11) in adults and children .net/policies-post/wma-declaration-of- Statistical Analysis who are at risk. helsinki-ethical-principles-for-medical- Statistical differences between groups Since 1989 the Belgian Diabetes Regis- research-involving-human-subjects/), and were assessed by x2 test, with Yates cor- try (BDR) has recruited first-degree rela- approved by the ethics committees of the rection or Fisher exact test for categorical tives who are ,40 years of age and BDR and the participating university hos- variables and Mann-Whitney U test for monitored their immunogenetic char- pitals. Random blood samples were col- continuous data. Kaplan-Meier analysis acteristics (IAA, GADA,IA-2A, zinc trans- lected for sera and buffy coats, and was used to estimate diabetes-free sur- porter 8 autoAbs [ZnT8A], and HLA-DQ aliquots were stored at 280°C until ana- vival. The survival curves were compared genotype) in relation to clinical outcome lyzed for diabetes-associated autoAbs using the log-rank test. Follow-up for without preselection according to initial and HLA-DQ genotype, respectively, as diabetes-free survival started at the time autoAb status or genotype (10). We pre- described previously (12). All relatives of the first autoAb+ sample and ended viously reported (12,13) that the pres- were followed regardless of their autoAb at the last contact with the relative or at ence of IA-2A or ZnT8A plus at least one status and HLA-DQ genotype. Relatives clinical onset, whichever came first. other diabetes autoAb type conferred a were not prescreened for islet cell cyto- When assessing multiple-autoAb positiv- high and age-independent risk of progres- plasmic autoAbs (ICA), nor were ICA re- ity as the end point, follow-up ended at sion to diabetes within 5–10 years in first- sults analyzed in the current study. thetimeofthefirst multiple-autoAb+ degree relatives. Here, we investigated AutoAb positivity was defined as being sample. The persistence of multiple- the influence of autoAb profile (number persistent if the next sample was also pos- autoAb positivity was defined as being still and type) and age at first autoAb positiv- itive for at least one autoAb regardless of positive for at least two autoAbs regard- ity on progression to type 1 diabetes in an the type. During follow-up, development less of the type at the next sampling time. extended group of siblings and offspring, of diabetes was ascertained through re- The following age strata were considered: who were followed over a longer obser- peated contacts of the BDR with Belgian 0–9, 10–19, and 20–39 years of age. The vation period than before (12,13). In endocrinologists and pediatricians, self- time to diabetes or to seroconversion to addition, we evaluated the impact of reporting through yearly questionnaires, multiple-autoAb positivity was assessed changes in autoAb profile during follow-up and a link with the BDR patient database, by Cox proportional hazards regression on long-term diabetes risk and the in- where newly diagnosed patients ,40 years analysis, or, in case later development fluence of HLA-DQ genotype on the de- of age are registered. Follow-up ended at of additional autoAbs was taken into ac- velopment of multiple autoAbs and the time of the last blood sampling or, in count, by time-dependent Cox regression diabetes. the case of progression to diabetes, at analysis. The baseline variables listed in care.diabetesjournals.org Gorus and Associates 1067

Supplementary Table 1 were first entered 20 years (Supplementary Fig. 2C). The Progression Rate According into a univariate Cox regression analysis characteristics at first autoAb positivity to HLA-DQ Genotype model for predicting the time to diabetes of relatives whose conditions progressed At baseline, the presence of HLA-DQ2 (with age as continuous variable) and to diabetes and those whose conditions was not significantly associated with those with a univariate P value ,0.05 did not (yet) progress during follow-up positivity for IAA or ZnT8A, but the sus- were then entered in a multivariate are shown in Supplementary Table 1. ceptibility haplotype was more prevalent 2 model. Two-tailed statistical tests were in GADA+ than in GADA relatives (54% performed using SPSS for Windows ver- Progression Rate According to AutoAb vs. 38%; P = 0.004); it tended to be less sion 24.0 (IBM SPSS Statistics, Chicago, Profile at Baseline prevalent in the presence of IA-2A than in IL). GraphPad Prism version 5.00 for Win- Multiple-autoAb+ relatives (n =194)pro- its absence (41% vs. 53%; P = 0.016) (data dows (GraphPad, San Diego, CA) was used gressed significantly faster to clinical on- not shown). In contrast, HLA-DQ8 was for the figures. set than relatives who were positive for strongly associated with the develop- only one autoAb at baseline (n = 268) ment of positivity for IA-2A and ZnT8A 2 RESULTS (progression after 15 years: 79% [95% CI (IA-2A+ vs. IA-2A : 82% vs. 12%; ZnT8A+ 2 Overall Progression to Diabetes 72–87%] vs. 30% [95% CI 22–39%]; P , vs. ZnT8A : 74% vs. 26%; P , 0.001 for Between March 1989 and December 0.001; progression after 20 years: 88% both) but not with the development of 2015 the BDR enrolled 7,029 siblings or [95% CI 80–95%] vs. 54% [95% CI 26– positivity for GADA or IAA (data not + + offspring (age range 0–39 years; 49% 87%]; P , 0.001) (Fig. 1A). In multiple- shown). IA-2A or ZnT8A relatives pre- female) of patients with type 1 diabetes autoAb+ relatives, there was no overall senting with positivity for at least one (see flowchart in Supplementary Fig. 1). significant difference in progression rate other autoAb carried HLA-DQ8 (both Four hundred sixty-two relatives (45% according to increasing numbers of HLA-DQ2/DQ8 and HLA-DQ8/non-DQ2 female) were identified as being persis- autoAb specificities present (P = 0.104) genotypes) more often than those pre- tently positive for IAA, GADA, IA-2A, (data not shown). The presence of IA-2A senting with positivity only for IAA and + and/or ZnT8A: 299 were autoAb at first or ZnT8A plus at least one additional GADA (78% vs. 56%; P = 0.002) or for a sampling, and 163 seroconverted to autoAb type among the four specificities single autoAb (78% vs. 48%; P = 0.001); in autoAb positivity during follow-up. In to- tested (n = 139) was associated with a contrast, the three groups did not differ in tal, diabetes developed in 171 (37%) per- frequency of HLA-DQ2 (50%, 59%, and + higher progression rate than in the case sistently autoAb relatives (40% female) of positivity for IAA and GADA alone 45%, respectively; P = 0.218) (data not during an overall median follow-up time shown). In relatives who were single (n =54,P = 0.001) (Fig. 1B). In the case + + of 59 months (interquartile range [IQR] of positivity for IA-2A or ZnT8A, the pro- autoAb (any type) or double autoAb 28–102 months) (129 initially autoAb+ rel- gression rate was not influenced by the for IAA and GADA at baseline, the pro- atives and 42 seroconverters after a me- gression rate to diabetes was strongly de- total number of other autoAbs present dian follow-up time since first autoAb pendent on the HLA-DQ genotype (Fig. (Fig. 1B); neither did it depend on IA-2A positivity of 63 months [IQR 29–109 1C)(P = 0.001), being highest in carriers of or ZnT8A levels (Supplementary Fig. 3A months] and 54 months [IQR 25–88 the HLA-DQ2/DQ8 genotype. In contrast, and B) or on whether only IA-2A, only months], respectively). There were 6,756 in multiple-autoAb+ relatives in whom 2 ZnT8A, or both were present (Supple- persistently autoAb or transiently IA-2A or ZnT8A had already developed, mentary Fig. 3C). The progression rate autoAb+ participants, among whom dia- the progression rate was no longer signif- of relatives who were double positive betes developed in 10 (median follow-up icantly different according to HLA-DQ for IAA and GADA was significantly higher time 86 months [IQR 73–121 months]; genotype (Fig. 1D)(P = 0.214), although than that of single-autoAb+ relatives (n = 15-year progression 0.3% [95% CI 0.1– there was still a nonsignificant trend to 268, P , 0.001) (Fig. 1B). 0.5%]). + more rapid progression in carriers of + In initially single-autoAb relatives, In the group of persistently autoAb HLA-DQ2/DQ8 genotype (P = 0.074 vs. GADA was the most prevalent autoAb relatives, the overall progression rate of all other genotypes) (data not shown). individuals who were autoAb+ at first (68% vs. 23% for IAA and 9% for IA-2A sampling was higher than that for partic- or ZnT8A), but the progression rate did Progression Rate According to ipants who were followed from the time not vary according to autoAb type Changes in AutoAb Profile of seroconversion (Supplementary Fig. 2A) (Supplementary Fig. 3D). It did, however, Initially, single-autoAb+ relatives in whom (P = 0.025). However, both groups did not increase with higher levels (stratified ac- positivity for additional autoAb types devel- differ in the time to onset of diabetes cording to tertiles) for GADA (P , 0.001) oped during follow-up (n = 71) progressed when compared after stratification accord- and IA-2A (P = 0.017), but not for IAA significantly faster to diabetes than those ing to age (data not shown). Neither did (P = 0.643) (data not shown). At baseline, who remained persistently positive they differ in the time from first multiple- IAA levels were inversely correlated and for a single autoAb (n = 197) (Fig. 2A); autoAb positivity to onset of diabetes GADA levels were positively correlated likewise, initially autoAb+ individuals (Supplementary Fig. 2B). Therefore, both with age (P =0.001andP =0.002,re- who became positive for IA-2A or ZnT8A groups were considered together for fur- spectively), in line with previous findings plus at least one other autoAb during ther analysis. The overall progression at the diagnosis of diabetes (4,6,17); lev- follow-up (n = 75) displayed a higher pro- (median) rates of all autoAb+ first-degree els of IA-2A and ZnT8A showed no signif- gression rate than relatives who never did relatives were 54% (95% CI 47–60%) after icant correlation with age (data not acquire this profile (n =246)(Fig.2B). 15 years and 70% (95% CI 55–85%) after shown). When considering the entire group of 1068 Twenty-Year Progression to Type 1 Diabetes Diabetes Care Volume 40, August 2017

multiple autoAbs (P = 0.001), and in car- riers of the HLA-DQ2/DQ8 genotype (P = 0.014); time to diabetes onset was longer with older age at baseline (P = 0.004) and in offspring of a mother with diabetes (P = 0.007) (Table 1, Model 1). The same exercisewasrepeatedinthegroupwith multiple-autoAb positivity (Table 1, Model 2). The presence of the high-risk autoAb profile was the only significant predictor of more rapid progression to diabetes in this setting (P , 0.001). In initially single-autoAb+ relatives (n = 268), Cox regression confirmed younger age (P , 0.001), the presence of HLA- DQ2/DQ8 genotype (P , 0.001), and pos- itivity for IAA (P = 0.028) as independent predictors of seroconversion to multiple- autoAb positivity (Supplementary Table 2, Model 1). A time-dependent Cox regression model for the prediction of diabetes identified seroconversion to multi- ple autoAbs as an independent risk fac- tor (P=0.012), in addition to younger age (P = 0.042) and HLA-DQ2/DQ8 carriership (P = 0.009) (Supplementary Table 2, Model 2). The loss of positivity for IAA dur- ing follow-up was not identified as an in- dependent risk factor in single-autoAb+ relatives, but it reached significance when performing time-dependent Cox analysis in the entire group of relatives (P = 0.002) (data not shown).

CONCLUSIONS By including persistently autoAb+ rela- Figure 1—Diabetes-free survival in persistently autoAb+ relatives stratified according to the number of autoAbs at baseline in the whole group (A), autoAb type and the number of autoAbs at baseline in tives over a large age range and following the whole group (B), HLA-DQ genotype in the absence of the high-risk (HR) autoAb profile (HR = them for a longer time than in our pre- IA-2A+ or ZnT8A+ plus at least one other autoAb+)atbaseline(C), and HLA-DQ genotype in the vious work (12,13), this study confirms presence of the high-risk autoAb profile at baseline (D). Log-rank test in panel B:*P = 0.001 vs. both that, in line with earlier findings in chil- fi , + + groups with a high-risk autoAb pro le (with n =2orn = 3 or 4 autoAbs); **P 0.001 vs. IAA and GADA dren followed from birth (3), the majority group. Because of a missing ZnT8A value at baseline, one relative (positive for IAA and GADA but not + for IA-2A) was excluded from analysis in panels B and D because of an uncertain initial risk profile. of multiple-autoAb individuals will prog- ress to the symptomatic disease phase relatives (n = 462), the loss of an autoAb first autoAb positivity in the group of rel- within 20 years and extends this observa- marker during follow-updin most in- atives who were positive for a single tion to adolescents and young adults. It fi stances IAA (n = 44) or GADA (n =31)d autoAb type or were double positive for con rms that multiple-autoAb positivity was associated with an overall decrease in IAA and GADA (Fig. 3A) but was age in- signals a high risk of progression to clinical progression rate (P = 0.016) (data not dependent in carriers of IA-2A or ZnT8A onset of diabetes (3,11), which occurs shown); this delay in progression was sig- presenting with at least one other autoAb more rapidly in the presence of IA-2A or nificant only for the loss of IAA (Fig. 2C) (Fig. 3B). Likewise, progression to diabe- ZnT8A regardless of age, HLA-DQ geno- (P = 0.014), but not for the loss of any of the tes of single-autoAb+ relatives, but not type, the number of additional autoAbs other specificities (Fig. 2D)(P = 0.307). In of multiple-autoAb+ relatives, decreased present, and the type of relationship to + particular, diabetes had not developed in with age (data not shown). the proband (12,13). In single-autoAb any of the 19 single-IAA+ relatives who relatives, younger age, the presence of became IAA2 during follow-up at their Multivariate Analysis HLA-DQ2/DQ8 genotype, and IAA positiv- + last visit (data not shown). In the entire group of autoAb relatives ity are associated with a higher propen- (n = 462), the time to diabetes onset was sity to seroconvert to multiple autoAbs. Progression Rate According to Age shorter in the case of the presence of When they reach that stage, they acquire The long-term rate of progression to di- IA-2A or ZnT8A plus at least one other the risk associated with this new profile. abetes decreased significantly with age at autoAb (P , 0.001), in the presence of Conversely, the loss of IAA positivity care.diabetesjournals.org Gorus and Associates 1069

ZnT8A at baseline progress more rapidly than those initially positive only for IAA and GADA, which is consistent with ob- servations that autoAb specificities de- velop sequentially, as a rule, with IA-2A and ZnT8A generally appearing closest to clinical onset (4,21). Relatives who are double positive for IAA and GADA progress more rapidly to disease onset than those who are single positive for an autoAb. In line with recent reports (22,23), the latter were most often GADA+. Their risk of progression to diabe- tes was independent of the autoAb type present but increased with higher levels (tertiles) of GADA or IA-2A. Using time- dependent Cox regression analysis, we found that the development of positivity for additional autoAbs was an indepen- dent determinant of progression to dia- betes in single-autoAb+ relatives, which is in agreement with previous reports (22–24) that seroconversion to multiple- autoAb positivity conferred a risk level similar to that for initially multiple- autoAb+ individuals. Such seroconversion occurred more often with younger age; the presence of HLA-DQ2/DQ8 genotype, similar to a previous study (22); and pos- itivity for IAA, a finding that is at variance with the same study (22). The loss of pos- itivity for IAAdbut not for GADA, IA-2A, d Figure 2—Diabetes-free survival in persistently autoAb+ relatives stratified according to the follow- or ZnT8A during follow-up tended to ing changes in autoAb profile during follow-up: later development of at least one additional autoAb delay the progression to diabetes, which in initially single-autoAb+ relatives (A); later development of the high-risk (HR) autoAb profile (HR = is in agreement with a recent report (25) IA-2A+ or ZnT8A+ plus at least one other autoAb+) in relatives without that profile at baseline (B); and is consistent with the proposed role later loss of initial IAA positivity in the whole group of relatives (C); and later loss of initial positivity of (pro)insulin autoimmunity in directing for GADA, IA-2A, or ZnT8A in the whole group of relatives (D). Because of a missing ZnT8A value at disease progression in many patients baseline, one relative (positive for IAA and GADA but not for IA-2A) was excluded from analysis in 2 panel B because of an uncertain initial risk profile. Subjects who lost positivity for IAA together with (25). In persistently autoAb or tran- that for another antibody (n = 7) were not included in panel D. Only one of these seven relatives siently autoAb+ relatives, the long-term progressed to diabetes. risk of diabetes was ,1%, which is in line with earlier studies (4,11). Our data on the associations among during follow-up is associated with a conclusions on the nearly complete 20-year autoAbs and genetic markers are in line lower risk. progression to diabetes in multiple-autoAb+ with the previously reported (17,26) clus- The strengths of our study include the relatives are limited by the low number tering at disease onset between GADA recruitment of participants among family of individuals followed for 20 years and positivity and HLA-DQ2/DR3, on the one members of a representative registry- the fact that not all relatives were fol- hand, and of IA-2A and HLA-DQ8/DR4,on based group of type 1 diabetes patients lowed from the time of seroconversion the other hand. The association of ZnT8A (2), the broad age range, and the lack of to autoAb positivity. However, when with HLA-DQ8, but not DQ2, is somewhat preselection according to HLA-DQ geno- stratified according to single- or multi- at variance with the association of ZnT8A type or ICA positivity. The measurement ple-autoAb positivity, seroconverters with both haplotypes in recent-onset di- of ZnT8A, which was shown to improve to autoAb positivity and initially autoAb+ abetes (17). The fact that the prevalence the prediction of the 5-year risk of type 1 relatives did not differ in progression of HLA-DQ8, but not of HLA-DQ2, was diabetes (11–13,18–20), also adds value rate. graded according to our autoAb-inferred to long-term risk assessment based on Our data confirm results obtained risk stratification (highest in relatives with positivity for IAA, GADA, and IA-2A (3). during a shorter follow-up period of positivity for multiple autoAbs carrying With two exceptions, the ZnT8A assay smaller groups that, regardless of the IA-2A and/or ZnT8A, intermediate if only could be performed in all samples posi- presence of two, three, or four autoAb positive for IAA and GADA, and lowest if tive for one or more other autoAbs. Our types, relatives positive for IA-2A and/or single autoAb+) is a novel finding. Our 1070 Twenty-Year Progression to Type 1 Diabetes Diabetes Care Volume 40, August 2017

function, insulin action, and glycemic con- Overall, childhood-onset type 1 diabe- trol (28,29). In addition, the relation to tes is characterized by a more aggressive HLA class I and non-HLA alleles, which disease course, more severe clinical pre- have been reported to influence the pro- sentation, positivity for more autoAbs, gression rate (30,31), needs to be ex- and a higher genetic risk load (6,34). How- plored further. This may allow the ever, regardless of age the presence of identification of episodes during which a IA-2A or ZnT8A plus at least one other large decrease in b-cell function is most autoAb enables us to identify individuals likely to occur; such episodes could be with a similar subclinical disease course, investigated further by measurements allowing first, if needed, the enrollment of of markers of b-cell death (32). Of note, adults afflictedbyadiseaseprocessrep- careful analysis of tissue from resentative of that in children, in terms autoAb+ organ donors revealed that in- of intensity and progression rate, in sec- sulitis could be observed only in the ondary prevention trials before the actual presence of multiple-autoAb positivity enrollment of children. Additional hor- at an asymptomatic disease stage monal/metabolic parameters should where b-cell mass appeared still to be then further secure the selection of par- unaffected (33). The proposed autoAb- ticipants in the same presymptomatic based risk stratificationcouldalsobe disease phase (11,28,29). Future work used to investigate episodes of b-cell should investigate whether our autoAb- loss after diagnosis. based prediction model may be improved

Table 1—Cox proportional hazards regression analysis for prediction of progression to diabetes based on baseline characteristics Univariate analysis Multivariate analysis* Hazard ratio Hazard ratio Covariates (95% CI) P value (95% CI) P value Model 1: overall study population Age 0.96 (0.94–0.98) <0.001 0.97 (0.95–0.99) 0.004 Sex, male vs. female (ref.) 1.26 (0.93–1.72) 0.135 NE BMI-SDS z score 1.06 (0.93–1.21) 0.413 NE HLA-DQ genotype DQ2/DQ8 2.09 (1.52–2.88) <0.001 1.51 (1.09–2.09) 0.014 DQ8/non-DQ2 1.02 (0.75–1.40) 0.904 NE — Figure 3 Diabetes-free survival in persis- DQ2/non-DQ8 0.78 (0.55–1.11) 0.165 NE + fi tently autoAb relatives strati ed according Non-DQ8/non-DQ2 0.34 (0.22–0.69) 0.001 NS 0.158 – – – to age (0 9, 10 19, and 20 39 years) in the Relationship with the proband absence (A)orpresence(B)ofhigh-risk(HR) Sibling 0.63 (1.20–2.22) 0.002 NS 0.527 autoAb profile (HR = IA-2A+ or ZnT8A+ plus at Offspring mother 0.33 (0.20–0.53) <0.001 0.51 (0.31–0.83) 0.007 least one other autoAb+). Because of a missing Offspring father 1.23 (0.88–1.71) 0.228 NE ZnT8A value at baseline, one relative (positive fi for IAA and GADA, but not for IA-2A) was Number/pro le of autoAbs – < – excluded from analysis in panel B because of Positivity for 2 to 4 autoAbs 5.22 (3.72 7.34) 0.001 2.27 (1.38 3.74) 0.001 fi – < – < an uncertain initial risk profile. HR pro le 5.03 (3.70 6.83) 0.001 2.38 (1.53 3.72) 0.001 Model 2: relatives with multivariate analyses showed that HLA- 2to4autoAbs** Age 0.98 (0.96–1.01) 0.136 NE DQ susceptibility haplotypes, in particular Sex, male vs. female (ref.) 0.99 (0.83–1.19) 0.980 NE when combined into the HLA-DQ2/DQ8 BMI-SDS z score 1.00 (0.86–1.16) 0.995 NE genotype, are important drivers of the HLA-DQ genotype development of positivity for multiple DQ2/DQ8 1.42 (0.98–2.06) 0.065 NE autoAbs; however, once that stage is DQ8/non-DQ2 0.73 (0.51–1.04) 0.084 NE reached, HLA class II genotype groups DQ2/non-DQ8 1.14 (0.75–1.75) 0.542 NE – seem to contribute little to the pro- Non-DQ8/non-DQ2 0.75 (0.35 1.61) 0.462 NE Relationship with the proband gression from advanced autoimmunity Sibling 1.30 (0.90–1.87) 0.158 NE to clinical onset of disease, which is in Offspring mother 0.50 (0.27–0.94) 0.030 0.059 line with the results of the study by Offspring father 1.05 (0.71–1.54) 0.813 NE Ilonen et al. (27). AutoAb profile It is now important to carefully investi- HR profile 2.23 (1.44–3.45) <0.001 2.23 (1.44–3.45) <0.001 gate further how the development of a HR, high autoantibody-inferred risk (IA-2A+ or ZnT8A+ plus positivity for one or more other autoAbs); NE, higher autoAb-inferred risk profile relates not selected for entry into the multivariate model. *Forward stepwise multiple regression model. to critical switches in the underlying dis- **One relative excluded from the analysis because the presence or absence of a high-risk profile at baseline could not be ascertained (missing ZnT8A value). Significant P values appearin boldface type. ease process and to changes in b-cell care.diabetesjournals.org Gorus and Associates 1071

by including autoAbs against newly iden- G. T’Sjoen, B. Lapauw, J. Ruige, A. Huytse, A. diabetic patients derives from seroconversion to tified autoantigens, such as tetraspanin-7 Rawoens, N. Steyaert, S. Deneve, and N. Platteau); persistent autoantibody positivity after 10 years – (35), or by replacing radiobinding assays and Universitair Ziekenhuis Leuven, Leuven, Belgium of age. Diabetologia 2012;55:413 420 (R. Bouillon, P. Gillard, M. Carpentier, M. Robijn, 5. Krolewski AS, Warram JH, Rand LI, Kahn CR. for established markers, in particular IAA K. Rouffe,´ A. Schoonis, H. Morobe,´ S. Achten, and Epidemiologic approach to the etiology of type I and GADA, by novel sensitive nonisotopic R. Van Heyste) for assisting in organizing the field diabetes mellitus and its complications. N Engl J assays claiming to recognize high-affinity work for sample and data collection throughout the Med 1987;317:1390–1398 autoAbs (36,37) or by assays measuring country. In addition, the authors thank all members 6. Gorus FK. Diabetes registries and early biolog- of the BDR who contributed to the recruitment of autoAbs against specific autoantigen frag- ical markers of insulin-dependent diabetes melli- relatives. F.K.G. thanks the JDRF Biomarker Working tus. Belgian Diabetes Registry. Diabetes Metab ments such as N-terminally truncated Group for useful discussions and suggestions. Rev 1997;13:247–274 GAD65 (38). Positivity for the latter two Funding. Funding was obtained for this project 7. Kyvik KO, Nystrom L, Gorus F, et al. The epide- types of assays has been reported to be from the Research Foundation–Flanders (Fonds miology of type 1 diabetes mellitus is not the – associated with a higher rate of progres- Wetenschappelijk Onderzoek [FWO] Vlaanderen same in young adults as in children. Diabetologia project nos. 3.9007.91, 7.0021.96, 3.0113.97, – sion to diabetes than that observed for 2004;47:377 384 G.0319.01, G.0514.04, G.0311.07, G.0374.08, 8. Bach JF, Chatenoud L. A historical view from conventional IAA or GADA radiobinding as- and G.0868.11 to F.K.G.), the Research Council thirty eventful years of immunotherapy in autoim- says (36–38). Further improvements may of the Vrije Universiteit Brussel (OZR grants mune diabetes. Semin Immunol 2011;23:174–181 come from analyzing autoAbs against spe- 1150, 1449, and 1650 to F.K.G., and Strategic Re- 9. Skyler JS. The compelling case for anti-CD3 in – cific autoantigen (neo)epitopes (39,40). search Program Growth grant SRP42 to F.K.G. type 1 diabetes. Diabetes 2013;62:3656–3657 and B.K.), the Willy Gepts Fund of Universitair 10. Gorus FK, Keymeulen B, Veld PA, Pipeleers In conclusion, in single-autoAb+ rel- Ziekenhuis Brussel (grants 3-2005 and 3/22-2007 DG. Predictors of progression to Type 1 diabetes: atives the development of additional to F.K.G. and grant G177 [2013] to I.W.), the Flemish preparing for immune interventions in the pre- immune markers occurs slower with in- Government (grant IWT 130138 to D.G.P.), and the clinical disease phase. Expert Rev Clin Immunol creasing age and the absence of IAA or European Union-Seventh Framework Programme 2013;9:1173–1183 HLA-DQ2/DQ8 genotype, but once the (FP-7 project no. 241883 to D.G.P.). F.K.G., C.M., 11. Insel RA, Dunne JL, Atkinson MA, et al. Staging B.K., and I.W. are/were recipients of a senior clinical presymptomatic type 1 diabetes: a scientific stage of multiple-autoAb positivity is research fellowship from FWO–Vlaanderen, statement of JDRF, the Endocrine Society, and reached, most individuals progress to and A.V.D. was the recipient of a junior research the American Diabetes Association. Diabetes the onset of diabetes within 20 years re- fellowship from the same institution. O.C. was Care 2015;38:1964–1974 gardless of age, HLA-DQ genotype, or the the recipient of a junior research fellowship 12. De Grijse J, Asanghanwa M, NoutheB,etal.;´ from the Agentschap voor Innovatie door number of autoAbs. Overall progression Belgian Diabetes Registry. Predictive power of Wetenschap en Technologie (IWT, Belgium). screening for antibodies against insulinoma- is more rapid if IA-2A or ZnT8A are present The funding organizations did not influence associated protein 2 beta (IA-2beta) and zinc and tends to be delayed in the case of a the data collection and interpretation. transporter-8 to select first-degree relatives of later loss of positivity for IAA. This autoAb- Duality of Interest. No potential conflicts of in- type 1 diabetic patients with risk of rapid pro- based risk stratification offers a platform to terest relevant to this article were reported. gression to clinical onset of the disease: impli- Author Contributions. All authors are mem- cations for prevention trials. Diabetologia 2010; further characterize and closely follow sub- bers of the BDR and/or its associated Reference b 53:517–524 groups in terms of functional -cell mass, Laboratory, and, as such, contributed to the data 13. Gorus FK, Balti EV, Vermeulen I, et al.; Belgian other biomarkers, and glycemic control, used in this article. F.K.G. analyzed the data and Diabetes Registry. Screening for insulinoma anti- both before and after the clinical onset wrote the manuscript. E.V.B. contributed to the gen 2 and zinc transporter 8 autoantibodies: a of diabetes. This is ultimately believed statistical support, analyzed the data, and wrote cost-effective and age-independent strategy to the manuscript. A.M. and I.W. analyzed the data, to advance our knowledge of the natural identify rapid progressors to clinical onset among contributed to the discussion, and reviewed and relatives of type 1 diabetic patients. Clin Exp Im- disease history and the underlying path- edited the manuscript. S.D., A.V.D., O.C., H.D., munol 2013;171:82–90 ological process and may contribute to the C.M., L.V.G., B.K., and D.G.P. contributed to the 14. American Diabetes Association. Classification tailoring of selection criteria for participa- discussion and reviewed and edited the manu- and diagnosis of diabetes mellitus. Sec. 2. In Stan- tion in prevention trials to the type of in- script. I.W. contributedto the statistical support. dards of Medical Care in Diabetesd2016. Diabe- F.K.G. and E.V.B. are the guarantors of this work tervention and the targeted disease stage. tes Care 2016;39(Suppl. 1):S13–S22 and, as such, had full access to all the data in the 15. Weets I, Van Autreve J, Van der Auwera BJ, study and take responsibility for the integrity of et al.; Belgian Diabetes Registry. Male-to-female the data and the accuracy of the data analysis. excess in diabetes diagnosed in early adulthood is Acknowledgments. The authors thank cowork- not specific for the immune-mediated form nor is ers at the central unit of the BDR, Brussels Free References it HLA-DQ restricted: possible relation to increased University-VUB,Brussels,Belgium(K.Verhaeghen, 1. Eisenbarth GS. Type I diabetes mellitus. A body mass index. Diabetologia 2001;44:40–47 V. Baeten, G. De Block, T. De Mesmaeker, chronic autoimmune disease. N Engl J Med 16. Van der Auwera BJ, Schuit FC, Weets I, Ivens H. Dewinter, N. Diependaele, S. Exterbille, T. 1986;314:1360–1368 A, Van Autreve JE, Gorus FK; Belgian Diabetes Glorieux, P. Goubert, C. Groven, T. Haulet, A. Ivens, 2. Vandewalle CL, Coeckelberghs MI, De Leeuw Registry. Relative and absolute HLA-DQA1-DQB1 D. Kesler, F. Lebleu, E. Quartier, G. Schoonjans, IH, et al. Epidemiology, clinical aspects, and biol- linked risk for developing type I diabetes before M. Van Molle, S. Vanderstraeten, T. Cluysen, J. De ogy of IDDM patients under age 40 years. Com- 40 years of age in the Belgian population: impli- Nayer, and A. Walgraeve) for expert technical parison of data from Antwerp with complete cations for future prevention studies. Hum Immu- assistance.Theauthorsalso thank their colleagues ascertainment with data from Belgium with 40% nol 2002;63:40–50 on the different university teams at Universitair ascertainment. The Belgian Diabetes Registry. Di- 17. Vermeulen I, Weets I, Asanghanwa M, et al.; Ziekenhuis Antwerpen, Antwerp, Belgium (I. De abetes Care 1997;20:1556–1561 Belgian Diabetes Registry. Contribution of anti- Leeuw,C.DeBlock,R.Braspenning,J.Michiels,and 3. Ziegler AG, Rewers M, Simell O, et al. Serocon- bodies against IA-2b and zinc transporter 8 to J. Vertommen); Universitair Ziekenhuis Brussel, version to multiple islet autoantibodies and risk of classification of diabetes diagnosed under 40 years Brussels, Belgium (G. Somers, K. Decochez, T. De progression to diabetes in children. JAMA 2013; of age. Diabetes Care 2011;34:1760–1765 Mesmaeker, S. Exterbille, P. Goubert, C. Groven, 309:2473–2479 18. Long AE, Gooneratne AT, Rokni S, Williams AJ, V. Kemels, C. Tettelin, S. Vanderstraeten, U. 4. Vermeulen I, Weets I, Costa O, et al.; Belgian Bingley PJ. The role of autoantibodies to zinc Van de Velde, and A. Walgraeve); Universitair Diabetes Registry. An important minority of transporter 8 in prediction of type 1 diabetes in Ziekenhuis Gent, Ghent, Belgium (R. Rottiers, prediabetic first-degree relatives of type 1 relatives: lessons from the European Nicotinamide 1072 Twenty-Year Progression to Type 1 Diabetes Diabetes Care Volume 40, August 2017

Diabetes Intervention Trial (ENDIT) cohort. J Clin 26. Genovese S, Bonfanti R, Bazzigaluppi E, et al. 33. In’t Veld P, Lievens D, De Grijse J, et al. Screen- Endocrinol Metab 2012;97:632–637 Association of IA-2 autoantibodies with HLA DR4 ing for insulitis in adult autoantibody-positive 19. Achenbach P, Lampasona V, Landherr U, et al. phenotypes in IDDM. Diabetologia 1996;39: organ donors. Diabetes 2007;56:2400–2404 Autoantibodies to zinc transporter 8 and SLC30A8 1223–1226 34. Wherrett DK, Chiang JL, Delamater AM, et al.; genotype stratify type 1 diabetes risk. Diabetolo- 27. Ilonen J, Kiviniemi M, Lempainen J, et al.; Type 1 Diabetes TrialNet Study Group. Defining gia 2009;52:1881–1888 Finnish Pediatric Diabetes Register. Genetic suscep- pathways for development of disease-modifying 20. Yu L, Boulware DC, Beam CA, et al.; Type 1 Di- tibility to type 1 diabetes in childhooddestimation therapies in children with type 1 diabetes: a con- abetes TrialNet Study Group. Zinc transporter-8 of HLA class II associated disease risk and class II sensus report. Diabetes Care 2015;38:1975–1985 autoantibodies improve prediction of type 1 diabe- effect in various phases of islet autoimmunity. 35. McLaughlin KA, Richardson CC, Ravishankar tes in relatives positive for the standard biochemical Pediatr Diabetes 2016;17(Suppl. 22):8–16 A, et al. Identification of tetraspanin-7 as a target autoantibodies. Diabetes Care 2012;35:1213–1218 28. Van Dalem A, Demeester S, Balti EV, et al.; of autoantibodies in type 1 diabetes. Diabetes 21. Yu L, Rewers M, Gianani R, et al. Antiislet Belgian Diabetes Registry. Relationship between 2016;65:1690–1698 autoantibodies usually develop sequentially glycaemic variability and hyperglycaemic clamp- 36. Fouts A, Pyle L, Yu L, et al.; Type 1 Diabetes rather than simultaneously. J Clin Endocrinol derived functional variables in (impending) type 1 TrialNet Study Group. Do electrochemilumines- Metab 1996;81:4264–4267 diabetes. Diabetologia 2015;58:2753–2764 cence assays improve prediction of time to diabe- 22. Bingley PJ, Boulware DC, Krischer JP; Type 1 29. Ferrannini E, Mari A, Nofrate V, Sosenko JM, tes in autoantibody-positive TrialNet subjects? Diabetes TrialNet Study Group. The implications Skyler JS; DPT-1 Study Group. Progression to di- Diabetes Care 2016;39:1738–1744 of autoantibodies to a single islet autoantigen to abetes in relatives of type 1 diabetic patients: 37. Mayr A, Schlosser M, Grober N, et al. GAD relatives with normal glucose tolerance. Diabeto- mechanisms and mode of onset. Diabetes 2010; autoantibody affinity and epitope specificity iden- logia 2016;59:542–549 59:679–685 tify distinct immunization profiles in children at 23. Xu P, Krischer JP; Type 1 Diabetes TrialNet 30. Mbunwe E, Van der Auwera BJ, Weets I, et al.; risk for type 1 diabetes. Diabetes 2007;56:1527– Study Group. Prognostic classification factors as- Belgian Diabetes Registry. In antibody-positive 1533 sociated with development of multiple autoanti- first-degree relatives of patients with type 1 di- 38. Williams AJK, Lampasona V, Wyatt R, et al. bodies, dysglycemia and type 1 diabetesda abetes, HLA-A*24 and HLA-B*18, but not HLA- Reactivity to N-terminally truncated GAD65 (96- recursive partitioning analysis. Diabetes Care B*39, are predictors of impending diabetes with 585) identifies GAD autoantibodies that are 2016;39:1036–1044 distinct HLA-DQ interactions. Diabetologia 2013; more closely associated with diabetes progression 24. Achenbach P, Warncke K, Reiter J, et al. 56:1964–1970 in relatives of patients with type 1 diabetes. Di- Type 1 diabetes risk assessment: improvement 31. Lempainen J, Laine AP, Hammais A, et al. abetes 2015;64:3247–3252 by follow-up measurements in young islet Non-HLA gene effects on the disease process of 39. Morran MP, Casu A, Arena VC, et al. Humoral autoantibody-positive relatives. Diabetologia type 1 diabetes: from HLA susceptibility to overt autoimmunity against the extracellular domain of 2006;49:2969–2976 disease. J Autoimmun 2015;61:45–53 the neuroendocrine autoantigen IA-2 heightens 25. Endesfelder D, Hagen M, Winkler C, et al. A 32. Herold KC, Usmani-Brown S, Ghazi T, et al.; the risk of type 1 diabetes. Endocrinology 2010; novel approach for the analysis of longitudinal pro- Type 1 Diabetes TrialNet Study Group. b cell 151:2528–2537 files reveals delayed progression to type 1 diabetes death and dysfunction during type 1 diabetes de- 40. Strollo R, Vinci C, Arshad MH, et al. Antibodies in a subgroup of multiple-islet-autoantibody-positive velopment in at-risk individuals. J Clin Invest 2015; to post-translationally modified insulin in type 1 children. Diabetologia 2016;59:2172–2180 125:1163–1173 diabetes. Diabetologia 2015;58:2851–2860